Measuring AI’s capability to accelerate biological research

Source: OpenAI Date: 2025-12-16 URL: https://openai.com/index/accelerating-biological-research-in-the-wet-lab

Summary

Title-only: OpenAI publishes a framework for measuring how much AI can accelerate biological research — evaluating AI’s capability to assist with experimental design, literature synthesis, hypothesis generation, and wet lab protocol development. The URL slug references “wet lab,” indicating this covers hands-on experimental biology (cell culture, protein assays, PCR) rather than purely computational biology. December 2025 places this alongside the GPT-5.2 series and the life sciences acceleration post.

Implications

The biological research acceleration thread. Measuring AI’s capability to accelerate biology is both a capability assessment and an existential dual-use question — the same capabilities that accelerate beneficial drug discovery can accelerate bioweapon development. OpenAI publishing this measurement framework signals they’re taking the biological acceleration question seriously enough to build evaluation infrastructure around it, not just rely on system card red-teaming.

Wet lab context. “Wet lab” acceleration is harder to measure than computational biology acceleration because wet lab experiments have physical constraints (reagent availability, incubation times, equipment access) that AI can advise on but not shortcut. Measuring how much AI accelerates wet lab researchers is therefore measuring something real about AI’s role as an expert assistant in physical science — distinct from AI doing computation on biological data.